rawhide hotplug broken

rawhide hotplug broken

[Jack Howarth]

   In case folks here are unaware, RedHat has been hacking
the hotplug srpm releases here for their rawhide releases
(and I assume their upcoming RH 7.1 release) doing such things
as removing the rc scripts from hotplug. On my linuxppc (running
Linux 2.4.4pre1, glibc 2.2.3 and the rawhide srpms), I find
that RedHat's modified hotplug doesn't work for loading the
usb print module when it sees my Epson 740i printer. Their
previous hotplug-1.9 release did work but RedHat claims this
was broken and loaded all modules blindly.
   I discovered last night that if I installed the latest 
hotplug-2001_02_28 and usbutils-0.7-101 rpms on my machine
that the usb print module was loaded on boot or hotplugging
the Epson 740i printer. I also discovered that I had to have
usbutils installed for the hotplugging to work. Also if I
substituted RedHat's rawhide hotplug-2001_02_14-15 rpm, the
hotplugging no longer worked with or without usbutils.
Perhaps some of the experts here should help out RedHat since
they seem to be a tad confused on how the hotplug stuff should
be set up. Oh, the bugzilla report for this problem is...

http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id5091

Thanks in advance for helping them out with this. 
                Jack




_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[Trond Eivind Glomsrød]

Jack Howarth <howarth@bromo.med.uc.edu> writes:

>    In case folks here are unaware, RedHat has been hacking
> the hotplug srpm releases here for their rawhide releases
> (and I assume their upcoming RH 7.1 release) doing such things
> as removing the rc scripts from hotplug. 

That script only initializes modules - we do this in rc.sysinit, so
USB keyboards can be used for interactive startup and if you have to
manually run things like fsck.

>    I discovered last night that if I installed the latest 
> hotplug-2001_02_28 and usbutils-0.7-101 rpms on my machine
> that the usb print module was loaded on boot or hotplugging
> the Epson 740i printer.

It does work with most other devices, I don't know why your printer is
an exception. You've tried an identical kernel with both versions of
hotplug? 

-- 
Trond Eivind Glomsrød
Red Hat, Inc.

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[Jack Howarth]

Trond,
   I am using Linux 2.2.4pre1 (from the linuxppc bitkeeper tree)
and find that the hotplugging works with the stock
hotplug-2001_02_28-1.src.rpm rpm from sourceforge. However it
requires usbutils to be installed. I find that your srpm in
rawhide for hotplug doesn't work with or without usbutils.
I'll look at the differences between your srpm and the stock
srpm to see if anything jumps out at me. I am a tad puzzled
why (if usbutils is essential for hotplugging to work in
the sourceforge release of hotplug) that you don't include
usbutils as part of rawhide.
                                  Jack

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[Bill Nottingham]

Jack Howarth (howarth@bromo.med.uc.edu) said: 
> hotplug-2001_02_28 and usbutils-0.7-101 rpms on my machine
> that the usb print module was loaded on boot or hotplugging
> the Epson 740i printer. I also discovered that I had to have
> usbutils installed for the hotplugging to work. Also if I
> substituted RedHat's rawhide hotplug-2001_02_14-15 rpm, the
> hotplugging no longer worked with or without usbutils.
> Perhaps some of the experts here should help out RedHat since
> they seem to be a tad confused on how the hotplug stuff should
> be set up.

Actually, from looking at this, it just appears to require newer
hotplug scripts than we ship. Or am I missing something?

Bill

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[David Brownell]

>    In case folks here are unaware, RedHat has been hacking
> the hotplug srpm releases here for their rawhide releases
> (and I assume their upcoming RH 7.1 release) doing such things
> as removing the rc scripts from hotplug.

It'd be interesting to know why that isn't being included.

Despite what Trond said (in a separate email on this thread), it
does more than just load modules; it also mounts "usbdevfs" at
the _right_ time (before USB host controller modules load) to
deal with some boot-time hotplugging ("cold-plugging") issues
that come up in some configurations ... like the way RedHat 7.0
initialized USB, or the last 7.1 init scripts I saw did.

Also, perhaps of more relevance to USB developers, "rc.usb"
provides some simple USB diagnostics ("rc.usb status") and
makes it easy to start, stop, and restart the USB subsystem.


>     On my linuxppc (running
> Linux 2.4.4pre1, glibc 2.2.3 and the rawhide srpms), I find
> that RedHat's modified hotplug doesn't work for loading the
> usb print module when it sees my Epson 740i printer. ...

Well, I did notice RedHat wasn't using current RPMs,
but I don't think that'd be the whole problem.


>    I discovered last night that if I installed the latest 
> hotplug-2001_02_28 and usbutils-0.7-101 rpms on my machine
> that the usb print module was loaded on boot or hotplugging
> the Epson 740i printer. I also discovered that I had to have
> usbutils installed for the hotplugging to work. 

Curious.  Can you try debugging the script invocation to see
why?  I usually:  (a) modify /sbin/hotplug to "env >/tmp/debug";
(b) insert the device, (c) modify a copy of that /tmp/debug file
and turn it into a script that exports all variables (or at least the
six used by USB hotplug :) and then does something like

    cd /etc/hotplug
    sh -x usb.agent

Then there's copious debug output (save it to a log file) which
can be examined ... in this case, to see why the "printer" driver
listing in /lib/modules/.../modules.usbmap doesn't match.

- Dave




_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[Bill Nottingham]

David Brownell (david-b@pacbell.net) said: 
> >    In case folks here are unaware, RedHat has been hacking
> > the hotplug srpm releases here for their rawhide releases
> > (and I assume their upcoming RH 7.1 release) doing such things
> > as removing the rc scripts from hotplug.
> 
> It'd be interesting to know why that isn't being included.

Since we initialize usb in rc.sysinit, it wouldn't do anything.

> Despite what Trond said (in a separate email on this thread), it
> does more than just load modules; it also mounts "usbdevfs" at
> the _right_ time (before USB host controller modules load)

We do that. The reason cold-plugging doesn't work with this because
the fs isn't read-write at the time. So in our current package,
hotplug actually waits around for a while for the root fs to
become read-write. It's not the most elegant solution, but
it was the quickest to code at the time.

Bill

_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: rawhide hotplug broken

[David Brownell]

> > Despite what Trond said (in a separate email on this thread), it
> > does more than just load modules; it also mounts "usbdevfs" at
> > the _right_ time (before USB host controller modules load)
> 
> We do that. 

Good to know -- the last version I looked at didn't work that way.

>    The reason cold-plugging doesn't work with this because
> the fs isn't read-write at the time. So in our current package,
> hotplug actually waits around for a while for the root fs to
> become read-write. It's not the most elegant solution, but
> it was the quickest to code at the time.

Hmm, that'd explain the original problem:

- Using hotplug scripts old enough that they demand
  that /tmp be writable ... so the pure bash version of
  hotplugging won't work;

- Not having "usbmodules", so the bash code is the
  only option.

Yes, that was an old bug in the hotplug scripts, now
fixed.

- Dave



_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel